Maximum Likelihood Data Fusion With Optimal Similarity Constraints for Full-Polarimetric RoPeR Imaging of Martian Regolith

The Zhurong rover of the Chinese Tianwen-1 mission is equipped with a Mars rover penetrating radar (RoPeR). The high-frequency CH-2 antenna adopts a full-polarimetric (FP) acquisition mode to map the fine structure of the Martian regolith. However, the data from different polarimetric channels characterize different features of the subsurface. The accurate interpretation of regolith structure relies on a clear and comprehensive imaging result. This article proposes a new data fusion algorithm to achieve this goal, which is called the maximum likelihood method with optimal similarity constraints (ML-OSC). A unique loss function is constructed to guarantee the similarity between the fusion result and source radargrams and simultaneously consider the amplitude difference in different channel data. FP ground penetrating radar (FP-GPR) experiments on three typical targets are performed in the laboratory. The comparisons with three normal methods validate the effectiveness and superiority of the proposed method. Finally, the proposed method is applied to the field FP-RoPeR data fusion imaging. Several subsurface structures are analyzed in detail to verify its effectiveness in Martian regolith imaging. The proposed method provides new and complete imaging results of Martian regolith, which will lead to a more accurate interpretation of it in the future.